Image and video synthesis has become a blooming topic in computer vision and machine learning communities along with the developments of deep generative models, due to its great academic and application value. Many researchers have been devoted to synthesizing high-fidelity human images as one of the most commonly seen object categories in daily lives, where a large number of studies are performed based on various deep generative models, task settings and applications. Thus, it is necessary to give a comprehensive overview on these variant methods on human image generation. In this paper, we divide human image generation techniques into three paradigms, i.e., data-driven methods, knowledge-guided methods and hybrid methods. For each route, the most representative models and the corresponding variants are presented, where the advantages and characteristics of different methods are summarized in terms of model architectures and input/output requirements. Besides, the main public human image datasets and evaluation metrics in the literature are also summarized. Furthermore, due to the wide application potentials, two typical downstream usages of synthesized human images are covered, i.e., data augmentation for person recognition tasks and virtual try-on for fashion customers. Finally, we discuss the challenges and potential directions of human image generation to shed light on future research.
translated by 谷歌翻译
Few-shot node classification is tasked to provide accurate predictions for nodes from novel classes with only few representative labeled nodes. This problem has drawn tremendous attention for its projection to prevailing real-world applications, such as product categorization for newly added commodity categories on an E-commerce platform with scarce records or diagnoses for rare diseases on a patient similarity graph. To tackle such challenging label scarcity issues in the non-Euclidean graph domain, meta-learning has become a successful and predominant paradigm. More recently, inspired by the development of graph self-supervised learning, transferring pretrained node embeddings for few-shot node classification could be a promising alternative to meta-learning but remains unexposed. In this work, we empirically demonstrate the potential of an alternative framework, \textit{Transductive Linear Probing}, that transfers pretrained node embeddings, which are learned from graph contrastive learning methods. We further extend the setting of few-shot node classification from standard fully supervised to a more realistic self-supervised setting, where meta-learning methods cannot be easily deployed due to the shortage of supervision from training classes. Surprisingly, even without any ground-truth labels, transductive linear probing with self-supervised graph contrastive pretraining can outperform the state-of-the-art fully supervised meta-learning based methods under the same protocol. We hope this work can shed new light on few-shot node classification problems and foster future research on learning from scarcely labeled instances on graphs.
translated by 谷歌翻译
深度学习(DL)通过前段光学相干断层扫描(AS-OCT)图像在角度闭合分类方面取得了重大进展。这些AS-OCT图像通常是通过不同的成像设备/条件获得的,这会导致基础数据分布的巨大变化(称为“数据域”)。此外,由于实用的标签困难,某些域(例如设备)可能没有任何数据标签。结果,在一个特定域(例如,特定设备)上训练的深层模型很难适应,因此在其他域(例如其他设备)上的性能很差。为了解决此问题,我们提出了一个多目标域的适应范式,以将在一个标记的源域上训练的模型转移到多个未标记的目标域。具体而言,我们提出了一种新型的多尺度多目标域对抗网络(M2DAN),以进行角度闭合分类。 M2DAN进行多域对抗性学习,以提取域不变特征,并开发一个多尺度模块,用于捕获AS-OCT图像的本地和全局信息。基于这些域不变的特征在不同尺度上,在源域上训练的深模型即使在这些域中没有任何注释,也能够在多个目标域上对角度闭合进行分类。对现实世界AS-OCT数据集进行的广泛实验证明了该方法的有效性。
translated by 谷歌翻译
本文研究了一个新的,实用但具有挑战性的问题,称为类无监督的域名适应性(CI-UDA),其中标记的源域包含所有类别,但是未标记的目标域中的类别依次增加。由于两个困难,这个问题具有挑战性。首先,源和目标标签集在每个时间步骤都不一致,这使得很难进行准确的域对齐。其次,以前的目标类在当前步骤中不可用,从而忘记了先前的知识。为了解决这个问题,我们提出了一种新型的原型引导连续适应(PROCA)方法,由两种解决方案策略组成。 1)标签原型识别:我们通过检测具有目标样本的累积预测概率的共享类来识别目标标签原型。 2)基于原型的对齐和重播:基于确定的标签原型,我们对齐域并强制执行模型以保留先前的知识。有了这两种策略,ProCA能够有效地将源模型改编为类未标记的目标域。广泛的实验证明了Proca在解决CI-UDA方面的有效性和优势。源代码可从https://github.com/hongbin98/proca.git获得
translated by 谷歌翻译
图形存在于许多现实世界中的应用中,例如财务欺诈检测,商业建议和社交网络分析。但是,鉴于图形注释或标记的高成本,我们面临严重的图形标签 - 刻度问题,即,图可能具有一些标记的节点。这样一个问题的一个例子是所谓的\ textit {少数弹性节点分类}。该问题的主要方法均依靠\ textit {情节元学习}。在这项工作中,我们通过提出一个基本问题来挑战现状,元学习是否是对几个弹性节点分类任务的必要条件。我们在标准的几杆节点分类设置下提出了一个新的简单框架,作为学习有效图形编码器的元学习的替代方法。该框架由有监督的图形对比学习以及新颖的数据增强,子图编码和图形上的多尺度对比度组成。在三个基准数据集(Corafull,Reddit,OGBN)上进行的广泛实验表明,新框架显着胜过基于最先进的元学习方法。
translated by 谷歌翻译
逐步学习新课程的能力对于所有现实世界的人工智能系统至关重要。像社交媒体,推荐系统,电子商务平台等的大部分高冲击应用都可以由图形模型表示。在本文中,我们调查了挑战但实际问题,图表几次拍摄的类增量(图形FCL)问题,其中图形模型是任务,以对新遇到的类和以前学习的类进行分类。为此目的,我们通过从基类循环地采样任务来提出图形伪增量学习范例,以便为我们的模型产生任意数量的培训集,以练习增量学习技能。此外,我们设计了一种基于分层的图形元学习框架,Hag-Meta。我们介绍了一个任务敏感的常规程序,从任务级关注和节点类原型计算,以缓解到新颖或基本类上的过度拟合。为了采用拓扑知识,我们添加了一个节点级注意模块来调整原型表示。我们的模型不仅达到了旧知识整合的更大稳定性,而且还可以获得对具有非常有限的数据样本的新知识的有利适应性。在三个现实世界数据集上进行广泛的实验,包括亚马逊服装,Reddit和DBLP,表明我们的框架与基线和其他相关最先进的方法相比,展示了显着的优势。
translated by 谷歌翻译
最近基于神经的关系提取方法虽然实现了对基准数据集的有希望的改进,但据对抗对抗攻击的脆弱性。到目前为止,努力主要集中在产生对抗性样本或捍卫对抗性攻击,但对正常和对抗样品之间的差异很少。在这项工作中,我们采取第一步利用基于显着的方法来分析那些对抗性样本。我们观察到显着标记与对抗扰动的直接相关。我们进一步发现对抗性扰动是训练集中不存在的那些代币或与关系标签相关的肤色提示。在某种程度上,我们的方法推出了对抗对抗样本的特征。我们在https://github.com/zjunlp/diagnoseadv中发布了一个开源测试用“diagnoseadv”。
translated by 谷歌翻译
更广泛的人重新识别(Reid)在最近的计算机视觉社区中引起了不断的关注。在这项工作中,我们在身份标签,特定特定因素(衣服/鞋子颜色等)和域特定因素(背景,观点等)之间构建结构因果模型。根据因果分析,我们提出了一种新颖的域不变表示,以获得概括的人重新识别(DIR-REID)框架。具体而言,我们首先建议解散特定于特定的和域特定的特征空间,我们提出了一种有效的算法实现,用于后台调整,基本上是朝向SCM的因果干预。已经进行了广泛的实验,表明Dir-Reid在大规模域泛化Reid基准上表现出最先进的方法。
translated by 谷歌翻译
在对象检测中,边界框回归(BBR)是决定对象定位性能的关键步骤。但是,我们发现BBR的大多数先前的损失功能都有两个主要缺点:(i)$ \ ell_n $ -norm和IOU基于IOU的损失功能都无法效率地描述BBR的目标,这会导致收敛速度缓慢和不准确的回归结果。 。 (ii)大多数损失函数都忽略了BBR中的不平衡问题,即与目标盒有较小重叠的大量锚盒对BBR的优化有最大的影响。为了减轻造成的不利影响,我们进行了彻底的研究,以利用本文中BBR损失的潜力。首先,提出了有关联合(EIOU)损失的有效交集,该交集明确测量了BBR中三个几何因素的差异,即重叠面积,中心点和侧面长度。之后,我们说明有效的示例挖掘(EEM)问题,并提出了焦点损失的回归版本,以使回归过程集中在高质量的锚点上。最后,将上述两个部分组合在一起以获得新的损失函数,即焦点损失。对合成数据集和真实数据集进行了广泛的实验。与其他BBR损失相比,在收敛速度和定位精度上都可以显着优势。
translated by 谷歌翻译
Knowledge graph embedding (KGE), which maps entities and relations in a knowledge graph into continuous vector spaces, has achieved great success in predicting missing links in knowledge graphs. However, knowledge graphs often contain incomplete triples that are difficult to inductively infer by KGEs. To address this challenge, we resort to analogical inference and propose a novel and general self-supervised framework AnKGE to enhance KGE models with analogical inference capability. We propose an analogical object retriever that retrieves appropriate analogical objects from entity-level, relation-level, and triple-level. And in AnKGE, we train an analogy function for each level of analogical inference with the original element embedding from a well-trained KGE model as input, which outputs the analogical object embedding. In order to combine inductive inference capability from the original KGE model and analogical inference capability enhanced by AnKGE, we interpolate the analogy score with the base model score and introduce the adaptive weights in the score function for prediction. Through extensive experiments on FB15k-237 and WN18RR datasets, we show that AnKGE achieves competitive results on link prediction task and well performs analogical inference.
translated by 谷歌翻译